Revealing driver-mediated indirect interactions between ecosystem services using Bayesian Belief Networks

Understanding the drivers mediating ecosystem service interactions is essential for supporting policy decisions aimed at sustaining synergies and mitigating trade-offs. Currently, most studies assessing ecosystem service interactions do not model them as a causal network. Here, we use Bayesian Belief Networks (BBNs) to assess how human activity intensity influences ecosystem service interactions (e.g., trade-off, synergy, no effect). We quantify changes in interactions for two snapshots in time in Southern Quebec (Canada) among aboveground forest carbon regulation, maple syrup provisioning, livestock provisioning, landscape recreation, bird-watching recreation, and number of bird species per route (an Essential Biodiversity Variable). By comparing correlation analyses to BBNs with or without the driver of human activity intensity, we show that not accounting for human activity intensity results in incorrectly attributing a driver-mediated trade-off as a direct trade-off (e.g., between bird-watching recreation and aboveground forest carbon regulation) and failure to detect direct interactions (e.g., between bird-watching recreation and livestock provisioning). BBNs provide a more complete understanding of interactions. In contrast to correlation analysis, which can only assess a relationship between two variables, BBNs can assess relationships among multiple variables and as such determine whether a relationship is due to a shared driver or whether the relationship is due to a direct synergy or trade-off among services. However, if relevant drivers are excluded, then direct interactions may be missed, and driver-mediated relationships may be incorrectly attributed as direct interactions. A better understanding of drivers that shape ecosystem service interactions could guide their management and provide targeted policy interventions.